Dean Phillips: America’s first AI candidate The Democratic challenger has sold out to Big Tech

There was nothing in Dean Phillips’s record as a nondescript Democratic backbencher from Minnesota to suggest that he had any special affinity for artificial intelligence: he entered Congress in 2019 as a campaign finance reform advocate opposed to the influence of big donor money. However, upon announcing his presidential campaign in October, he began to be seen around wealthy and influential figures in the tech world. As a political novice with no national base or constituency, it perhaps only made sense for Phillips to seek the patronage of new benefactors, who could potentially make a big difference in improving the odds of his long-shot run. Soon, he was describing Sam Altman, CEO of OpenAI, as “an extraordinary thinker and friend and voice of counsel”, while figures including Elon Musk, Andrew Yang and Bill Ackman have expressed their support, with Ackman even donating $1 million to his campaign. Phillips, the populist foe of plutocracy, suddenly became its sympathetic friend and admirer, prompting one former ally to point out that it was “a disgrace to see how far he’s fallen”.

In November, these increasingly cosy connections with Silicon Valley yielded a shift in the direction of the Phillips campaign. We Deserve Better, a Super PAC headed by Matt Krisiloff (formerly Altman’s chief of staff) and now funded in part by Ackman himself, was founded to aid the candidate, who began to sound exceedingly bullish about AI and cryptocurrency — two subjects which Phillips had previously had little to say about, but which are near and dear to many of the lords of Silicon Valley. In December, Phillips appeared at an event called “Stand with Crypto”, in which he claimed that AI and crypto should be allowed to flourish without too much regulation, echoing the libertarian ethos that predominates in the Valley. He said: “We should make sure we don’t stifle innovation, that we don’t stifle decentralisation when it is thoughtful and supportive of our national interest … that’s true in crypto as it should be in AI.” At another event in which Phillips was introduced by Yang, he boldly predicted that he would be “the first AI president in American history!”

Perhaps in an attempt to manifest this, in the middle of this month, Krisiloff’s Super PAC launched “Dean.bot”, a chatbot built with the help of the AI startup Delphi that could speak to voters and answer questions with Phillips’s voice and mannerisms. It was constructed using samples drawn from his speeches and interview appearances, and differed from Delphi’s previous political AI voice applications, which could only respond with pre-written answers. In other words, this was the first candidate bot one could have a free-flowing, real-time conversation with.

However, there was a problem: the chatbot was powered by ChatGPT, the LLM software owned by OpenAI, which prohibits its use for political campaigns. Delphi’s account was accordingly suspended by OpenAI and the developer has since taken it down. Krisiloff has signalled that other open-source software may be utilised to maintain the chatbot’s linguistic and interactive abilities, while tech researchers warned against its potential for spreading misinformation through the malicious impersonation of candidates. In any event, Dean.bot remains offline, for now.

It may seem insignificant, a novel experiment in AI that lasted all of a few days, but a Rubicon has surely been crossed and Dean.bot’s fleetingly short lifespan nonetheless offers a preview of what AI is capable of when brought into the political arena. As the demise of Dean.bot illustrated, the cultural climate is not yet ready to accord free rein to the use of this kind of technology in this context. But the fact is that such applications are here and they have proven to be easily deployable: it’s arguably only a matter of time, perhaps a few election cycles from now, when resistance to it eventually wears down and the experiment is repeated by campaigns in the future. It’s not difficult to imagine OpenAI or other firms announcing that, after much deliberation, they have determined the right set of standards to govern the use of AI in politics before lifting their prohibitions. From then on, the introduction of AI, like the introduction of social media a generation ago, would be treated as both desirable and inevitable.

After all, the American political class has often displayed an overeager attitude to the adoption of new technologies, borne of the Whiggish faith in unfettered progress that underlies their culture. And has ensured their all too ready deference to the wishes of the tech industry. As Phillips himself claims: “I come from a Congress in which… we have very few people who even understand it. And if you do not understand it, all you want to do is regulate it…” This is another way of saying that those who made the technology are the ones best suited to regulate it and government should, therefore, stay out of the way.

It’s a stance that mirrors the establishment’s initially fawning and effectively laissez-faire approach to social media in the Obama era, before the same technologies were blamed for destabilising American democracy in 2016 and became objects of intense suspicion. Today, the cycle seems to be repeating itself with AI: multiple rounds of hype and optimism at the early stages to be followed by an existential reckoning later on. Despite recognition of its ills, however, social media hasn’t perceptibly changed from what it was before. Likewise, it’s anyone’s guess if public authorities, as currently constituted, will even have the capacity, or, indeed, the political will, to redirect AI in any meaningful way.

For their part, Sam Altman and other leading Silicon Valley figures have espoused maximalist positions on the possibilities of open-ended AI development, even to the point of welcoming the emergence of new forms of consciousness. As Altman wrote in a 2017 essay speculating about “The Merge” of human and post-human intelligence: “It’s going to get a lot weirder. We will be the first species ever to design our own descendants. My guess is that we can either be the biological bootloader for digital intelligence and then fade into an evolutionary tree branch, or we can figure out what a successful merge looks like.”

These philosophical currents resemble the tenets of what has come to be known as “effective accelerationism” or “e/acc”, an influential ideology in tech circles arising from the writings of “Dark Enlightenment” thinker Nick Land. “Nothing human makes it out of the near future,” Land wrote, with his followers concluding that AI in particular and capitalist progress in general are forms of non-human super-intelligence that will eventually consume and supersede humanity itself. And though Altman doesn’t openly identify as an accelerationist, he and others from OpenAI, such as Krisiloff, have already done much to bring Land’s vision closer to fruition. Through their engagement with political figures including Phillips, such ideas and sentiments may soon find receptive outlets in the political mainstream.

In the not-so-distant future, when AI candidate bots are accepted as a normal feature of politics, the prospect of an “AI president” thus becomes much more plausible, though perhaps in ways radically different from what Phillips intended: less The West Wing, and more Black Mirror. And it is in this context of technology’s unexpected consequences that experiments like Dean.bot ought to be evaluated. For if Dean Phillips is the first “accelerationist candidate”, he certainly won’t be the last.